Skip to main content
. 2022 Oct 18;9:1011071. doi: 10.3389/fcvm.2022.1011071

Figure 2.

Figure 2

Predictive performance increased with multi-rather than single-cohort training data both in CoxNet (A) and survival gradient boosting (SGB) (B). On the pooled dataset, SBG achieved a greater overall c-index than CoxNet (C). In terms of number required features to achieve maximum predictive performance, SBG required 11 features and CoxNet 6 features (D).